Prediction of Long-Term Failure in Kevlar 49 Composites

Creep rupture data in Kevlar 49 epoxy usually exhibit considerable scatter: the coefficient of variation (CV) about the mean failure time at a given stress exceeds 100%. Quasi-static strength data, in contrast, shows little scatter: <4% CV for pressure vessels and <10% for impregnated strands. In this paper analysis of existing creep rupture data on Kevlar epoxy vessels at four storage pressures has produced an interesting and useful result. It was found that a significant portion of the scatter in failure times for pressure vessels is due to spool-to-spool variation in the eight spools of Kevlar fibers used to wind the vessels. The order rank of mean times to failure was consistent over a pressure range from 3400 to 4300 psi, 68 to 86% of short term burst. Also, the coefficient of variation about the mean failure time for each spool was less than that for the total sample. The statistical inference that the sample is nonhomogeneous was supported by a nonparametric check using the Kruskal-Wallis test, and by a parametric analysis of variance. The order rank found in long-term tests did not unequivocally agree with static strength ranks; several spool sets were distinctly high or low. The implication ismore » that, while static strengths are not valid predictors of long-term behavior, short term creep rupture tests at high stress definitely are. The material difference which causes the spool-to-spool variations has not yet been identified for all eight spools. However, it appears that Kevlar behavior at lower pressures may be predicted through the use of curves fitted to the data for each spool. A power law relating failure time to pressure, t = t/sub 0/(p/p/sub 0/)/sup m/, was found to fit the data reasonably well. The implication is that, both in composite vessel design and in creep rupture experiments, the pressure (or stress) level be carefully controlled.« less